← back

m625 is not a clinical decision tool · not a publisher · not an expert authority · not a simulator. It is a cross-domain Roll-formalized evidence atlas with public audit + attack surface.
persona: researcher · domain: cancer · tier: 1
ferry target: m590 · fixture id: DEMO-015
self-audit grade: A

i_demo_015 · researcher × cancer · m610 cooked-case re-audit invitation

Why a cancer-research methodologist would come to evidence.x1000.ai

A researcher studying cancer dynamics methodologically — specifically, the question "how does a Roll-formalized cooked case differ from the typical case-report in clinical-research literature" — needs a concrete worked example of the audit pipeline and a path to re-audit. Pain points:

m625 does not produce new cancer-research findings. **m625 invites cancer-research methodologists to re-audit m610's 3 base cooked cases under m610's own SELF_AUDIT_BENCHMARK_v0.1 (10 rules × 4 bindings) and m610's MATH_AUDIT pipeline, and to file findings via m625's challenge intake**.

This is not clinical-research advice. This is not an evaluation of any specific paper or sponsor.


The Roll-formalized object: the 3 base cooked cases as audit-replication candidates

(K4.1 bit pattern per m604 V2.6 sovereign reference · Kernel §3.4 · Framework §4.4-4.5; m610 F1-F5 audit families per m604 V2.6 Framework §4.1-4.5.)

CaseTMU typeK4.1m610 manual / machine audit
i_case_cancer_001 (EGFR-NSCLC)TMU-A primary · cytostatic(0, 1, 1)92/100 manual · 97/100 machine (v0.5.1 calibration ±1)
i_case_cancer_150 (BRCA1/2 breast)TMU-A synthetic-lethality(1, 1, 0)97/100 (post-fix · F3 hard-veto fix history is itself the v0.1→v0.2 lesson)
i_case_cancer_450 (R/R DLBCL CD19 CAR-T)TMU-C-major + TMU-A-minor(1, 1, 1) phi_direct_rewrite92/100 manual · 91→98/100 machine

All audit scores are training_data_recall · TODO_unverified (sourced from 01_materials_cache/A3_m610_cooked_cases_anchor.md). Live re-audit is the invitation.


How a researcher methodologist should approach (≤ 5 minutes)


What this demo will not say (anti-drift)


Cross-domain transferability declaration (W6.3)

The 3 cooked cases span 3 different K4.1 bit patterns: (0,1,1) cytostatic · (1,1,0) synthetic-lethality · (1,1,1) phi_direct_rewrite. Cross-domain transferability of the audit methodology (not the specific cases) to other domains (neuro · longevity · CKD) is via m620 verify ladder. transferability_pct: bit-pattern necessary, substantive m620 verdict sufficient.


Falsifying prediction

The "m610 cooked-case audit is replicable" claim is falsified by:


Attack surface entries


Ferry target

m590 (claim_pack · for graph anchoring) · m620 (verify_req if researcher requests live re-audit).


Honest verdict (W6.5)

- *"This is just an invitation, not evidence."* — Correct. The framework's value is the audit-replication procedure, not new evidence. - *"What if no researcher takes you up on this?"* — Then the framework correctly remains untested by external audit. v0.3+ public BLOCKED log will reflect that honestly.